Xiaoqing Luo

Heartbeat classification using disease-specific feature selection

Automatic heartbeat classification is an important technique to assist doctors to identify ectopic heartbeats in long-term Holter recording. In this paper, we introduce a novel disease-specific feature selection method which consists of a one-versus-one (OvO) features ranking stage and a feature search stage wrapped in the same OvO-rule support vector machine (SVM) binary classifier. The proposed method differs from traditional approaches in that it focuses on the selection of effective feature subsets for distinguishing a class from others by making OvO comparison. The electrocardiograms (ECG) from the MIT-BIH arrhythmia database (MIT-BIH-AR) are used to evaluate the proposed feature selection method. The ECG features adopted include inter-beat and intra-beat intervals, amplitude morphology, area morphology and morphological distance. Following the recommendation of the Advancement of Medical Instrumentation (AAMI), all the heartbeat samples of MIT-BIH-AR are grouped into four classes, namely, normal or bundle branch block (N), supraventricular ectopic (S), ventricular ectopic (V) and fusion of ventricular and normal (F). The division of training and testing data complies with the inter-patient schema. Experimental results show that the average classification accuracy of the proposed feature selection method is 86.66%, outperforming those methods without feature selection. The sensitivities for the classes N, S, V and F are 88.94%, 79.06%, 85.48% and 93.81% respectively, and the corresponding positive predictive values are 98.98%, 35.98%, 92.75% and 13.74% respectively. In terms of geometric means of sensitivity and positive predictivity, the proposed method also demonstrates better performance than other state-of-the-art feature selection methods.

Distance metric learning for soft subspace clustering in composite kernel space

Soft subspace clustering algorithms have been successfully used for high dimensional data in recent years. However, the existing algorithms often utilize only one distance function to evaluate the distance between data items on each feature, which cannot deal with datasets with complex inner structures. In this paper, a composite kernel space (CKS) is constructed based on a set of basis kernels and a novel framework of soft subspace clustering is proposed by integrating distance metric learning in the CKS. Two soft subspace clustering algorithms, i.e., entropy weighting fuzzy clustering in CKS for kernel space (CKS-EWFC-K) and feature space (CKS-EWFC-F) are thus developed. In both algorithms, the prototype in the feature space is mapped into the CKS by multiple simultaneous mappings, one mapping for each cluster, which is distinct from existing kernel-based clustering algorithms. By evaluating the distance on each feature in the CKS, both CKS-EWFC-K and CKS-EWFC-F learn the distance function adaptively during the clustering process. Experimental results have demonstrated that the proposed algorithms in general outperform classical clustering algorithms and are immune to ineffective kernels and irrelevant features in soft subspace. The composite kernel space is constructed based on a set of basis kernels.The general form of soft subspace clustering in CKS is presented.CKS-EWFC-K and CKS-EWFC-F are proposed under the framework of CKS-SSC.The properties of CKS-EWFC-K and CKS-EWFC-F are investigated.Both CKS-EWFC-K and CKS-EWFC-F are immune to ineffective kernels.

Regional and Entropy component analysis based remote sensing images fusion

In this paper a new regional and entropy component analysis based fusion approach is proposed for multispectral and panchromatic images fusion. The input images are decomposed into low frequency subband and high frequency subbands by stationary wavelet transform. The low frequency subband coefficients of multispectral image are selected as those of fused image. The fused rule of high frequency subbands are designed according to the similarity of corresponding region. The similar corresponding regions are fused by magnitude maximum rule, otherwise, fused by statistical model method. In order to obtain the corresponding region, the input images are divided into windows and extracted the integrate features from panchromatic and multispectral images. Inspired by the kernel entropy component analysis, the linear entropy component analysis ECA is proposed and used to extract the spectral feature. Different from traditional regional fusion approaches dividing input images separately, ours is generated from the synthetic features. The region result can be gotten by feature clustering using Fuzzy C-means, which is mapped into each of high frequency subbands. Some experiments are taken on some remote sensing images, and the results show the superiorities of our proposed method, both in subjective evaluation and some numerical guidelines.

Image fusion with contextual statistical similarity and nonsubsampled shearlet transform

Image fusion has the capability to integrate useful information from source images into a more comprehensive image. How to obtain the effective representation of source images is a key step to image fusion. Due to the loss of the dependence of coefficients, most of traditional multi-scale decomposition-based image fusion methods suffer from an inaccurate image representation. To solve this problem, a novel image fusion method with contextual statistical similarity in nonsubsampled shearlet transform (NSST) is presented. The key contributions include: 1) the dependence of NSST coefficients is captured by the contextual hidden Markov model (CHMM); 2) the contextual statistical similarity of coefficients is proposed; 3) an effective fusion rule based on the characteristic of CHMM is developed for high-frequency subbands in NSST domain. By the visual analysis and quantitative evaluations on experimental results, the superiority of the proposed method is demonstrated.

Adaptive multistrategy image fusion method

Abstract. Design of fusion rule is an important step in fusion process. Traditional single fusion rules are inflexible when they are being used to fuse feature-rich images. To address this problem, an adaptive multistrategy image fusion method is proposed. Its flexibility lies in the combination of a choose-max strategy and a weighted average strategy. Moreover, the region-based characteristics and the shift-invariant shearlet transform (SIST)-based activity measures are proposed to guide the selection of strategies. The key points of our method are: (1) Window-based features are extracted from the source images. (2) Use of the fuzzy c-means clustering algorithm to construct a region map in the feature difference space. (3) The dissimilarity between corresponding regions is employed to quantify the characteristic of regions and the local average variance of the SIST coefficients are considered as activity measures to evaluate the salience of the related coefficient. (4) The adaptive multistrategy selection scheme is achieved by a sigmoid function. Experimental results show that the proposed method is superior to the conventional image fusion methods both in subjective and objective evaluations.

Image Fusion Using Region Segmentation and Sigmoid Function

In this paper, a novel region segmentation and sigmoid function based image fusion method is proposed. Different from the traditional fusion approaches limiting to a single fusion strategy, the proposed method is designed with an adaptive multi-strategy fusion rule (AMFR). In our method, the source images are decomposed into low frequency sub bands and high frequency sub bands via the shift-invariant Shear let transform (SIST). The low frequency sub bands are fused by the choose-max scheme and the high frequency sub bands are fused by the AMFR based on a sigmoid function. The AMFR includes the choose-max scheme and the weighted average scheme, which of them is selected is determined by the sigmoid function. The fused sub bands are merged to reconstruct fused image by using inverse SIST. Experiments conducted on various types of source images demonstrate that our approach achieve superior results compared with the existing fusion methods in both visual presentation and objective evaluation.

Statistical Modeling of Multi-modal Medical Image Fusion Method Using C-CHMM and M-PCNN

In this paper, a new Contextual hidden Markov Model (CHMM) and modified Pulse Coupled Neural Network (M-PCNN) based fusion approach in the Contour domain is proposed for multi-modal medical image fusion. The Contour transform as an emerging multi-scale multi-direction geometric analyzing tool can provide an efficient and flexible representation of images, e.g. edges, contours and textures, which overcomes the drawback of the 2-D wavelet transform. Considering the powerful advantages for statistical modeling and processing of Contour let coefficients by HMM, the context information integrated with HMM is established to construct a comprehensive statistical correlative model, which can collectively capture persistence across scales, directional selectivity within scales and energy concentration in the spatial neighborhood of the high-frequency sub-band coefficients. Low-frequency sub-band coefficients are fused by the magnitude maximum rule, and a modified PCNN is developed where the linking strength of each neuron is determined by the normalized region energy of Edge PDF and modified spatial frequency is employed as the image feature to motivate M-PCNN. The high-frequency directional sub-band coefficients are selected by total pulse number maximum strategy. The experimental results demonstrate that the presented fusion method can further improve fusion image quality and visual effects.

Multi-focus image fusion using HOSVD and edge intensity

The edge intensity metric is proposed.The relationship between the patch energy and the shrink factor of the sigmoid function is constructed.Multi-strategy fusion rule based on edge intensity and patch energy is designed.The proposed fusion method is performed in the HOSVD domain. The purpose of multi-focus image fusion is to integrate the partially focused images into one single image which is focused everywhere. To achieve this purpose, higher order singular value decomposition (HOSVD) and edge intensity (EDI) based multi-focus image fusion method is proposed. The main characteristics of the proposed method includes: 1. an effective and robust sharpness measure based on edge intensity is presented; 2. considering the fact that HOSVD is an effective data-driven decomposition technique and shows the outstanding ability in the representation of high-dimensional data, it is used to decompose multi-focus images; and 3. a multi-strategy fusion rule based on sigmoid function is used to fuse the decomposition coefficients. Furthermore, several experiments are conducted to verify the superiority of the proposed fusion framework in terms of visual and statistical analyses.

Scalable learning method for feedforward neural networks using minimal-enclosing-ball approximation

Training feedforward neural networks (FNNs) is one of the most critical issues in FNNs studies. However, most FNNs training methods cannot be directly applied for very large datasets because they have high computational and space complexity. In order to tackle this problem, the CCMEB (Center-Constrained Minimum Enclosing Ball) problem in hidden feature space of FNN is discussed and a novel learning algorithm called HFSR-GCVM (hidden-feature-space regression using generalized core vector machine) is developed accordingly. In HFSR-GCVM, a novel learning criterion using L2-norm penalty-based ε-insensitive function is formulated and the parameters in the hidden nodes are generated randomly independent of the training sets. Moreover, the learning of parameters in its output layer is proved equivalent to a special CCMEB problem in FNN hidden feature space. As most CCMEB approximation based machine learning algorithms, the proposed HFSR-GCVM training algorithm has the following merits: The maximal training time of the HFSR-GCVM training is linear with the size of training datasets and the maximal space consumption is independent of the size of training datasets. The experiments on regression tasks confirm the above conclusions.

A New Pan-Sharpening Method Using Statistical Model and Shearlet Transform

ABSTRACT A statistical model-based pan-sharpening method under the framework of Shearlet transform is developed in this paper. The approximation coefficients of multi-spectral (MS) image are used for reconstruction level of fusion algorithm. The weighted average method is used to fuse the high-pass subbands. The weights are estimated by a statistical model in which a new objective function is proposed. It is partitioned into two parts: the first term maximizes the local variance of the high-pass subband of fused image, which means the details can be injected into MS image as much as possible. The second term is the correlation restriction between the high-pass subband of fused image and that of MS image, which is indirectly beneficial to avoid generating the unreasonable high-frequency information. Finally, the pan-sharpening result is obtained by performing the inverse Shearlet transform. A series of experiments conducted on real remote sensing images demonstrate the effectiveness of our model.